Environment configuration

Conda virtual env

The preferred option to setup your environment is through conda environment.

The setup of environment can be done in two steps. Navigate to the root of the downloaded repository, then

  1. Create a conda environment:

conda env create -f env/conda_env.yml

or

conda env create -f env/conda_gpu_env.yml

if you want to install PyTorch with cuda support.

  1. Activate the conda environment:

conda activate supernnova

or

conda activate supernnova-cuda

if you create environment from “conda_gpu_env.yml”.

  1. A python project management tool poetry (2.x) is installed via the above steps. Verify it and install python dependencies for this project:

which poetry # should print <conda env>/bin/poetry
poetry install

For developers (including testing local documentation), please refer to Python environments for development.

Mise

You can also use mise, a tool to manage multiple python environments.

  • Make sure mise is activated in your shell

eval "$(mise activate bash)"   # or zsh

  • Install Python 3.11 + uv

mise install

  • Install SuperNNova + dev deps

mise run install

  • Install PyTorch (pick one depending if you want cuda support or not)

mise run install-torch          # CPU
mise run install-torch-cuda     # CUDA 12.1

  • Run tests

mise run test

Docker

You can also use docker. The Docker image uses uv to install Python 3.11 and all dependencies (instead of conda), so there are no Anaconda Terms-of-Service prompts during the build.

Create a docker image:

make {image}

where image is one of cpu or gpu (for the latest supported CUDA version; currently 12.3.1) or gpu9 (for cuda 9.0)

  • This image contains all of this repository’s dependencies.

  • Image construction will typically take a few minutes

Enter docker environment by calling:

python env/launch_docker.py --image <image> --dump_dir </path/to/data>

  • Add --image image where image is cpu or gpu (latest version) or gpu9 (for cuda 9)

  • Add --dump_dir /path/to/data to mount the folder where you stored the data (see Data walkthrough) into the container. If unspecified, will use the default location (i.e. snndump)

This will launch an interactive session in the docker container, with zsh support.